Proceedings of the 2002 Winter Simulation Conference

E. Yücesan, C.-H. Chen, J. L. Snowdon, and J. M. Charnes, eds.

PROJECT MANAGEMENT – CRITICAL PATH METHOD (CPM)

AND PERT SIMULATED WITH PROCESSMODEL

LeRoy F. Simmons

Information Systems and Operations Management

Loyola College
4501 N. Charles St.
Baltimore, MD 21210, U.S.A.

ABSTRACT

ProcessModel simulation allows the activity times of a pro-

ject to be represented by a variety of distributions and fur-
ther the resulting project time may also be represented by a
variety of distributions. This is a significant improvement
over the traditional methods of CPM and PERT. Program
Evaluation and Review Technique (PERT) takes the CPM
network and adds distributions to represent the activity
times of the project. CPM assumes the activity times to be
constant, which is not likely in the real world. PERT as- Figure 1: Network of Project
sumes the activity times of the project to be distributed as
Beta distributions and the resulting project time to be a
Normal distribution. This is better than assuming them to
be constant, but these assumptions are needlessly restric-
tive. This paper demonstrates how simulation with Proc-
essModel can remove these needless restrictions.

1 EXAMPLE FOR CPM

Chase et al.(2001 on page 68) present a project for design-

ing a computer. The project requires the seven activities in Figure 2: Results from ProcessModel
Table 1 be completed.
The “Design” activity (Node1) starts the project and
Table 1: CPM Example finishes after 21 weeks. At that point the “Build Proto-
Immediate Time type” (Node2) and “Evaluate equipment”(Node3) activities
Activity Designation Node # Predecessors (weeks) begin. The Node2 activity ends after week 26 and at that
Design A 1 - 21 point the “Test prototype” (Node4) activity begins. …
Build prototype B 2 A 5
Evaluate equipment C 3 A 7
The project is completed in 38 weeks. Since the
Test prototype D 4 B 2 “Write equipment report” (Node5) activity ends without
Write equipment report E 5 C,D 5 another activity beginning, this activity is said to have
Write methods report F 6 C,D 8 “slack” time. The “Write equipment report” activity could
Write final report G 7 E,F 2
take another 3 weeks and the project would not be delayed.
All the other activities are said to be “critical”, because
The network in Figure 1 was developed using the Note
they have no slack time. If a critical activity takes longer
#’s and the Immediate Predecessors.
to complete than expected, the project will also take longer
Simulating this network with ProcessModel provided
to complete.
Figure 2.

1786
 Simmons

Once the above network is modeled with Process- Table 3: Output from ProcessModel
Model it is easily taken to PERT and beyond by simply
adding distributions to the model.

2 EXAMPLE FOR PERT

Chase et al. (2001 on page 71) suggested using the data in

Table 2 to perform a PERT analysis.

Table 2: PERT Example

Time
Activity Designation Node # optimistic most-likely pessimistic
Design A 1 10 22 28
Build prototype B 2 4 4 10
Evaluate equipment C 3 4 6 14 3 CPM AND PERT USING PROCESS MODEL
Test prototype D 4 1 2 3
Write equipment report E 5 1 5 9 The modeling trick that enables CPM/PERT networks to
Write methods report F 6 7 8 9 be modeled in ProcessModel involves the following of two
Write final report G 7 2 2 2
rules.
The following graph (Figure 3) was generated in Excel
• Rule 1 – When route(s) [arc(s)] exit a Node, there
from the output from a ProcessModel simulation. The
must be one “100%” route. If there is more than
graph shows the distribution of the project completion
one route, the others are made “Create” routes.
time.
In Figure 4 when Node1 is finished Nodes 2-
4 may begin.
• Rule 2 – When a Node has more that one route
coming into it, “Batching” is used at that Node,
where the batch size is the number of routes com-
ing into the Node.
In Figure 5 Node5 will not begin until all of it
immediate predecessors (Node2, Node3, and
Node4) are finished.

Figure 3: Completion-Time Graph

Chase et al. (2001) show how the traditional approach

to PERT analysis would calculate the probability of the
project being completed in 35 weeks. Their result was
that there is a 19% chance the project will be completed in
35 weeks. The results from the ProcessModel simulation Figure 4: Implementation of Rule 1
suggest that 15% would be a better estimate. The tradi-
tional approach estimated the project-time variance to be
11.90 while ProcessModel’s estimate was 8.35. The tradi-
tional approach has overestimated the variability of the dis-
tribution. Further, the following output (Table 3) from
ProcessModel suggests that the project completion time is
better estimated with distributions other than the Normal
distribution.
How does one get ProcessModel to perform a
CPM/PERT-like analysis?

Figure 5: Implementation of Rule 2

1787
 Simmons

The following template will get you started. Free

download of the template is available at (http://
leroysimmons.com).
The template (Figure 6) has all the ProcessModel
structure that is needed to perform a CPM/PERT for up to
twenty activities (Nodes). If more activities are required
additional Nodes can be added.

Figure 8: End Node

Figure 9: Start Node

N ode
N ode8 N ode9
10

Figure 6: ProcessModel Template

Each Node is preprogrammed as in Figure 7. The

variable graph01 is set to 1 when Node1 activity begins N ode3 N ode6

and then it is return to 0 when it ends. When this variable Ite m

is later graphed it will represent the occurrence of the S ta rt N ode1 N ode7 E nd

Node1 activity. Node2 is preprogrammed with graph02 Item 3

=2, Node3 is preprogrammed with graph03=3, and so on. N ode2 N ode4 N ode5
The variable Node1_Time for Node1 allows the activity Item 2

time to be entered elsewhere. Node2 has been prepro-

Item 4
grammed Node2_Time, Node3 has been preprogrammed
Node3_Time, and so on.
N ode N ode N ode N ode N ode N ode N ode N ode N ode N ode
11 12 13 14 15 16 17 18 19 20

Figure 10: Chase Example Using Template

Free download of this completed model is available at

(http://leroysimmons.com).

Figure 7: Preprogrammed Nodes
The end activity (Figure 8) is a dummy activity that
ends the project thus allowing the project time (variable –
ProjTime) to be recorded.
Stored in the Start activity (Figure 9) are all the activ-
ity times. As seen below the CPM/PERT template has all
the activity times by default set to zero. By replacing the
zeros with constants you will have the ProcessModel ver-
sion of CPM. By replacing the zeros with distributions
you will have the ProcessModel version of PERT.
Figure 10 shows what the Chase example would look
like using the CPM/PERT template.
REFERENCE
Chase, Richard, Nicholas Aquilano and Robert Jacobs,
2001, Operations Management for Competitive Ad-
vantage, ninth edition, McGraw-Hill.
AUTHOR BIOGRAPHY
LEROY F. SIMMONS is a Professor of Information Sys-
tems and Operations Management at Loyola College of
Maryland. He received his Ph.D. from the University of
Tennessee in 1973. His research interests are in business
forecasting and the simulation of business’ processes.

1788